Sam Stevens

[Home] [Writing] [Research] [Contact] [CV]

[About] [News] [Research]

About Me

My name is Sam Stevens. I’m a Ph.D. student at THE Ohio State University, where I work in computer science, specifically in AI-accelerated data-driven scientific discovery, with my advisor Prof. Yu Su.

I won Best Student Paper at CVPR 2024, gave the opening keynote at Biocuration 2025 and develop an open-source sparse-autoencoder library.

My research includes work in interpretability, computer vision, AI for cryptography and various LLM projects.

I did my Bachelor’s of Science in Computer Science at Ohio State (and minored in German). I’ve been lucky to intern at Zoom, Meta, SpaceX, Microsoft and GE Aviation (twice). I also studied abroad in Dresden for eight weeks.

I also previously worked on TicketBay with Salty Software.

News

  • 04/2025: Gave the opening keynote at the 18th International Biocuration Conferece in Kansas City. I discussed Imageomics projects including BioCLIP and SAEs for vision. The recording is now available.
  • 04/2025: Released a preprint about the small-data regime in modern AI evaluation: no recent popular AI methods papers evaluate on benchmarks with between 100 and 1K training samples.
  • 02/2025: Released saev which applies sparse autoencoders to vision models.
  • 06/2024: BioCLIP won Best Student Paper at CVPR 2024!
  • 05/2024: Interning with Zoom remotely over the summer.
  • 03/2024: Visited Tsukuba University in Japan to present recent work in ML and AI.
  • 03/2024: Gave a talk on BioCLIP at Science n Suds at Parsons North Brewing Company.
  • 02/2024: BioCLIP accepted to CVPR 2024!
  • 12/2023: Released BioCLIP, a foundation vision model for the entire tree of life.
  • 11/2023: Happy to see MMMU publicly released. I am super proud to be part of such an important benchmark for LMMs!
  • 05/2023: Released SELM, our work on symmetric encryption using language models, on arXiv.
  • 05/2023: Started as a Research Scientist Intern at Meta AI in Seattle, working with Kristin Lauter and Francois Charton on using Transformers to attack post-quantum cryptography algorithms.
  • 01/2022: I traveled with the Imageomics crew to Kenya for three weeks to broaden my understanding of science and gather data for future work.
  • 06/2022: OSU placed 3rd in Amazon’s Alexa TaskBot competition (as team TacoBot)! This is the first ever TaskBot competition; it was a fantastic experience working with applied NLP in such a competitive environment. More coverage here and here. Check out our website for more information!
  • 11/2021: Placed 4th at Hack OHI/O (as team “Killer Food Robots”) with an app to find the optimal trick-or-treating route Umar Jara and with two first time hackers, Blake Morse and Sam Latshaw!
  • 08/2021: Paper on pre-trained language model interpretability accepted to the EMNLP 2021 workshop BlackboxNLP!
  • 05/2021: Internship at SpaceX in Seattle, working on the Starlink team!
  • 11/2020: Won Best UI/UX and Buckeye’s Choice Awards at Hack OHI/O (as team //todo) with an app to convert voice to code using BART, a custom natural language to code neural model and a gorgeous React app by Garrett Morse. Video demo here.
  • 05/2020: Internship at Microsoft (remotely)! Working on the Power BI team.
  • 11/2019: Won Hack OHI/O for the 2nd time with an app to make text more accessible with OCR and text-to-speech.
  • 06/2019: Starting study abroad program in Dresden, Germany!
  • 05/2019: Awarded the Huntington International Fellowship!
  • 04/2019: Honorable mention for best visualization at DataFest 2019!
  • 01/2019: College of Engineering published an article about TicketBay!
  • 05/2018: Internship at GE Aviation! Working in GE Digital on end-to-end testing frameworks.
  • 11/2017: Won Hack OHI/O with an app to define and show examples of trending terms on the Internet.
  • 06/2017: Internship at GE Aviation through Cincinnati’s INTERAlliance program.

Research

Some selected publications are below. A list of all my papers is here. You can also check out my Google Scholar for a more up-to-date list.

Sparse Autoencoders for Scientifically Rigorous Interpretation of Vision Models
Samuel Stevens, Wei-Lun Chao, Tanya Berger-Wolf, Yu Su (arXiv Preprint) [paper] [website] [demos] [models]

BioCLIP: A Vision Foundation Model for the Tree of Life
Samuel Stevens*, Jiaman Wu*, Matthew J Thompson, Elizabeth G Campolongo, Chan Hee Song, David Edward Carlyn, Li Dong, Wasila M Dahdul, Charles Stewart, Tanya Berger-Wolf, Wei-Lun Chao, Yu Su (* equal contribution) (CVPR 2024, Best Student Paper) [paper] [website] [demo]

Memorization for Good: Encryption with Autoregressive Language Models
Samuel Stevens, Yu Su. (arXiv Preprint) [paper] [website] [code]

An Investigation of Language Model Interpretability via Sentence Editing
Samuel Stevens, Yu Su. (EMNLP BlackboxNLP 2021.) [paper] [code]

Getting Started in ML/AI

I often am asked how to get started in machine learning and artificial intelligence. I recommend starting with Coursera’s machine learning course1 and Andrej Karpathy’s Neural Networks: Zero to Hero. Both courses are very high-quality and should provide a lot of value compared to other free resources.

If you are a student getting started and feel lost, feel free to email. I also consult for AI/ML (and general software) if you are looking for professional services.

Non-Research Projects

TicketBay: fellow OSU students and I developed a mobile app for Ohio State University students to sell second-hand football tickets. (Jan 2018 - Present)

Quiet HN: a simple Hacker News site with no comments.


  1. Coursera offers other machine learning courses, including deep learning specializations. I haven’t taken theses courses, so I cannot comment on their quality.↩︎


[Links] [Source]

Sam Stevens, 2024